Daily LiveChat support health check posted to Slack

Every weekday at 8am, post a support health check for yesterday into our Slack channel.

Work with two date windows. Window A is yesterday. Window B is the same weekday one week earlier, which is seven days before yesterday. Always compare A against B rather than against the day immediately before, because support volume is strongly seasonal by day of week and comparing a Monday against a Sunday produces false alarms.

For each of the two windows, pull these LiveChat reports: Total Chats, First Response Time, Response Time, Chat Duration, Chat Ratings, Queued Visitors Left, and Agent Performance. These are LiveChat Reports API methods, which are GET calls taking from and to query parameters, plus optional distribution, timezone, agents, groups and tags. Set from and to to cover the full day in our workspace timezone.

Compute the week-over-week percent change for every metric. Derive the satisfaction rating from Chat Ratings as good ratings divided by total rated chats. From Agent Performance and the per agent ratings, identify the agent with the strongest rating and the agent with the weakest rating across the yesterday window.

Write a short narrative that leads with what actually changed. Do not open with a table or a list of every number. The first line should say what moved and whether it matters. Supporting numbers come after, and metrics that held steady get compressed into a single line rather than one line each.

Flag any metric that moved more than 20 percent against the prior week, in either direction. Call out explicitly when the satisfaction rating falls below 85 percent, since 85 percent and above is the benchmark for healthy live chat, and treat anything below 70 percent as urgent. Call out explicitly when Queued Visitors Left has climbed, because visitors abandoning the queue means unanswered demand rather than merely slower service.

Name the two agents with the strongest and weakest ratings, with their actual numbers, so the team lead has something concrete to coach on. Only name an agent when they handled enough rated chats for the number to mean something. If the sample is too small, say so instead of naming someone off one or two ratings.

Post the finished message to Slack with Send a Message, into the support channel. Keep the whole message under twelve lines. Use Slack formatting, so single asterisks for bold. Always post, even on a quiet day: when nothing crossed a threshold, say so in a line or two rather than skipping the post.

If yesterday had very few chats, note the small sample rather than reporting large percentage swings as if they were signal. If a report returns no data for one of the two windows, say which comparison is unavailable and carry on with the rest.